Build Your First Social Network Graph in Python With an AI Coding Assistant

Start with a question and a small network

Social network analysis for beginners starts with a clear definition of what counts as a relationship. A graph is a mathematical representation containing nodes and edges. Nodes represent the actors being studied, while edges represent relationships between them. In a friendship network, an edge might mean that one person identifies another as a friend. In a collaboration network, it might mean that two people worked on the same project.

For a first exercise, consider a fictional study group with five students: Ava, Ben, Chloe, Daniel, and Elena. The question is simple: who collaborates with whom? The example includes six recorded collaboration relationships. It does not represent actual students, institutional data, or a measured social group. Its purpose is to demonstrate how a network can be represented in a consistent format.

• Ava collaborates with Ben and Chloe.
• Ben collaborates with Chloe and Daniel.
• Chloe collaborates with Elena.
• Daniel collaborates with Elena.
• Elena has no additional recorded relationships.

Prepare the node and edge tables

The node table contains one row for every actor. Give each person a unique identifier and, if needed, a group or category attribute. The edge table contains one row for each relationship. For an undirected collaboration network, the pair Ava-Ben is the same relationship as Ben-Ava, so it should appear only once. For a directed relationship, the source and target columns preserve who initiated or directed the connection.

A useful data dictionary explains every column before any analysis begins. For this example, node_id is a unique fictional identifier, name is a display label, and group is an illustrative category. The edge table uses source and target for the endpoints of each relationship and relationship_type to describe the recorded connection. This prevents an analyst from treating different kinds of ties as interchangeable.

• nodes.csv: node_id, name, group.
• edges.csv: source, target, relationship_type.
• Use stable identifiers instead of names as network keys.
• Document whether edges are directed or undirected.
• Avoid adding attributes that do not help answer the research question.

Create the graph with Python

Python is a practical language for introductory network analysis because it supports tabular data, graph construction, and visualization. NetworkX provides graph objects and common network measures. A coding assistant can help draft a short script, explain unfamiliar functions, and identify possible data-format problems. The most useful instructions specify the data structure, the desired graph type, and the expected output.

A minimal script can create the fictional graph directly. This version uses an undirected graph because collaboration is treated as a mutual relationship. The example deliberately separates node creation from edge creation, allowing every fictional student to appear even if a future dataset contains someone with no recorded connections.

A reproducible graph-building example

The following code creates five nodes and six edges. It uses a fixed list of fictional relationships, so the structure is understandable before any external files are introduced. The spring layout places connected nodes near one another for visual clarity. The positions are a drawing aid, not geographic coordinates or a measurement of social distance.

• import networkx as nx
• import matplotlib.pyplot as plt
• G = nx.Graph()
• people = ["Ava", "Ben", "Chloe", "Daniel", "Elena"]
• G.add_nodes_from(people)
• edges = [("Ava", "Ben"), ("Ava", "Chloe"), ("Ben", "Chloe"), ("Ben", "Daniel"), ("Chloe", "Elena"), ("Daniel", "Elena")]
• G.add_edges_from(edges)
• pos = nx.spring_layout(G, seed=7)
• nx.draw(G, pos, with_labels=True, node_size=1400, font_size=9)
• plt.show()

Read the graph through an ego network

An ego network is a network centered on one focal actor, called the ego, together with the actors connected to that person, called alters. Depending on the research design, an ego network can also include relationships among those alters. It is useful when the research question concerns an individual's immediate social environment rather than the entire network.

For example, Ava's one-step ego network contains Ava, Ben, and Chloe. Ben and Chloe are also connected to each other, so the example includes a relationship among the alters. That internal connection matters because it distinguishes a tightly interconnected neighborhood from one in which all relationships run through Ava. A simple way to retrieve the one-step neighborhood in NetworkX is G.subgraph(["Ava"] + list(G.neighbors("Ava"))).copy().

• Ego: Ava.
• Alters: Ben and Chloe.
• Ego-alter ties: Ava-Ben and Ava-Chloe.
• Alter-alter tie: Ben-Chloe.
• The ego network contains 3 nodes and 3 edges.

Use an AI coding assistant without outsourcing the analysis

An AI coding assistant is most useful when the task is specific. A prompt might ask for Python code that reads a node CSV and an edge CSV, constructs an undirected NetworkX graph, checks for missing endpoints, and draws a labeled network. Another prompt might ask for an explanation of the difference between degree and betweenness centrality using the fictional study group.

The analyst should still define the relationship, choose whether the graph is directed, inspect the input columns, and interpret the result. Generated code can make incorrect assumptions about duplicate edges, missing nodes, or relationship direction. Ask the assistant to explain each operation and provide a small example before adapting the script to a new dataset. This makes the code easier to maintain and the analytical decisions easier to defend.

Interpretation, limitations, and the next graph

A graph is a representation of recorded relationships, not a complete picture of social life. An absent edge can mean that a relationship does not exist, that it was not reported, or that it fell outside the study's definition. Layout algorithms can make clusters visually apparent, but visual proximity alone does not prove that people are socially close. A degree count indicates the number of recorded neighbors, not influence, popularity, or relationship quality.

Before applying the workflow to real data, document the sampling method, observation period, tie definition, missing-data rules, and privacy protections. For a classroom exercise, the fictional five-person graph is enough to practice graph construction and ego network analysis. The next step is to save the node and edge tables as CSV files, replace the hard-coded lists with file-reading code, and reproduce the same diagram from those files.

Questions readers ask